The Google Update No One Confirmed
Large sites crashed on September 4, fully recovered on September 13, and Google never confirmed a thing. Volatility is spiking again this week, and the nine-day reversal is the best argument yet for touching nothing while it runs.
If your rankings fell off a cliff around September 4 and came completely back on September 13, you did not imagine it, and you did not fix it. Glenn Gabe, who tracks large-site impact with first-party analytics, posted on September 13 that “a number of the sites that fell off a cliff around 9/4 completely surged back yesterday. They dropped heavily in Search and some in Discover as well. The recovery is wild.” Barry Schwartz of Search Engine Roundtable verified the important part: these were not third-party visibility artifacts but “a real drop and then a real reversal in its first-party analytics and data.” Google confirmed nothing, on the way down or on the way back up.
Then the story restarted. On September 15, Schwartz reported a huge spike in chatter: UK results churning, spam on page one, keyword sets swapping to entirely different pages, one webmaster describing “that kind of change [that] happens only on big updates.” I checked Semrush Sensor myself this afternoon (September 16): US desktop, all categories, reads 4.1 out of 10, up from roughly half a point on September 13. Sensor calls 4.1 “normal range,” but three days ago it read half a point, and it has been climbing since.
None of this is comfortable for people paid to react to rankings, but the last nine days say the most valuable action during an unconfirmed volatility event is often no action at all, because Google sometimes undoes the event for you. Sites that rewrote content, disavowed links, or shipped “recovery” work between September 4 and September 12 spent that effort against a change Google rolled back on its own.
What the last month looked like
September’s whiplash makes more sense inside the run of events since mid-August. Google confirmed exactly one of them.
As of today, Google’s Search Status Dashboard shows no ranking incident for September 2026. Google’s confirmation protocol covers core updates and spam updates. Everything else, including whatever moved on September 4 and got reverted on September 13, happens in silence.
Why the volatility trackers disagreed about September 4
Three trackers covered the same week and gave different accounts of it.
Semrush Sensor
DataForSEO Volatility Index
For a real, first-party-verified crash to stay invisible to one major tracker, the hit has to be narrow. The most plausible reading, and it is inference rather than a Google statement, is that the September 4 event hit a specific cluster of large sites, hard, in Search and Discover. Aggregate trackers sample tens of thousands of SERPs across categories, and a sharp hit to even a few hundred properties barely registers in that average, especially under a freshly recalibrated scoring scale. A real, damaging event can read as calm on an aggregate score.
For practitioners, that reframes the familiar advice. “Sensor is calm, you’re fine” was never sound, and September proved it in both directions. If your own analytics show a cliff and the trackers show nothing, believe your own analytics. An aggregate score can dilute a real, narrow hit until it looks like nothing.
Google has rolled things back before
Gabe’s September 13 post carried one easy-to-miss clause: “And note, this has happened in the past.” It has, and the record goes back a decade. None of these reversions were announced.
| When | What happened | How it was known |
|---|---|---|
| Nov 2015 | Large unexplained ranking shifts; Google responded that there was no update to announce. | Schwartz, SER |
| Jun 2017 | Core-ranking shift that then “rolled back or kind of adjusted” within days. | Schwartz, SER |
| Mar 2019 | Core update that functioned as a partial rollback; Sistrix data showed losers of prior updates regaining. | Search Engine Journal |
| Jun 2020 | What looked like “a 10–20% rollback or adjustment of the May 5th core update,” unconfirmed by Google. | Schwartz, SER |
| Aug 2020 | An unconfirmed update “rolled back after a few hours”; rankings reverted to their prior state. | Schwartz, SER |
| Jun 2021 | Google itself warned that June core update changes “may see a reversal or shift” with the July update. A rare announced-in-advance reversion. | Schwartz, SER |
| Dec 2023 | Weeks of extreme volatility after the reviews update; observers concluded Google “rolled back the November core update or at least part of it.” | Schwartz, SER |
| Sep 2026 | Sites crash Sep 3–4 in Search and Discover, fully reverse Sep 13, verified in first-party analytics. No confirmation either direction. | Gabe via Schwartz, SER |
Confirmed updates are announced on the status dashboard, roll out over days or weeks, and stick. Unconfirmed events behave like experiments: a fast hit to particular site types, then a quiet reversion within one to two weeks in a meaningful share of cases. September 4 to September 13 is nine days, almost exactly the window the August 2020 rollback needed in miniature and the November 2023 partial rollback needed at full scale.
The push-revert loop
Google has never published a postmortem of the September event. The loop below is inference, built on the company’s observable behavior and Patel’s August 31 statement that the spam team “keeps updating” its detection systems.
Triage for this week’s spike
Volatility is live right now: Sensor at 4.1 and rising, UK SERPs churning, Discover unstable, and Schwartz expecting no confirmation either way.
The reports behind the current spike are geographically lumpy: “UK results are absolutely crazy since Saturday night. A lot of spam on page 1,” one webmaster wrote in Schwartz’s September 15 thread, while US trackers barely moved at first. Regional churn during EU DMA SERP redesigns is a known confound this month. The same thread also contains a shop owner reporting that conversions dried up at “4 PM on the dot” while traffic held, which smells like tracking or attribution breakage rather than a ranking change, so the sensible move is to verify before acting.
Two classes of volatility
Google’s ranking system now changes fast enough that its own operators sometimes reverse course within days, and its communication policy has not changed to match. The status dashboard covers updates Google wants to own, and it will never cover the ones Google decides to take back.
Confirmed updates are durable, announced, and worth restructuring for. Unconfirmed volatility is real while it lasts, frequently self-corrects, and punishes overreaction. The sites that came back on September 13 did nothing for nine days, and that nothing was the correct strategy. If the current spike follows the same script, the sites that make no panicked changes through early October will be the ones still standing when the chart flattens.
Primary sources: Search Engine Roundtable, “Google Search Ranking Update Around September 4th That Reverted September 13th” (Sep 14, 2026); Search Engine Roundtable, “Google Search Ranking Volatility Heats Up This Morning” (Sep 15, 2026); Glenn Gabe’s September 12–13 posts on X and Bluesky; Search Engine Watch, “Could Google’s next core update arrive in September 2026?” (Sep 10, 2026); Semrush Sensor, captured by the author Sep 16, 2026; Search Engine Roundtable, “Volatility Continues Post August Spam Update” (Aug 26, 2026); Search Engine Roundtable on Rajan Patel’s statement (Aug 31, 2026).
